Advances in Water Resources 29 (2006) 974–986
www.elsevier.com/locate/advwatres
Storage of water on vegetation under simulated rainfall
of varying intensity
R.F. Keim
a
a,*
, A.E. Skaugset b, M. Weiler
c
School of Renewable Natural Resources, Louisiana State University, LSU AgCenter, 227 Renewable Natural Resources Building,
Baton Rouge, LA 70803, USA
b
Department of Forest Engineering, Oregon State University, 215 Peavy Hall, Corvallis, OR 97331, USA
c
Faculty of Forestry, University of British Columbia, 2424 Main Mall, Vancouver, BC, Canada V6T1Z4
Received 6 December 2004; received in revised form 19 July 2005; accepted 25 July 2005
Available online 13 October 2005
Abstract
Little is understood about how storage of water on forest canopies varies during rainfall, even though storage changes intensity
of throughfall and thus affects a variety of hydrological processes. In this study, laboratory rainfall simulation experiments using
varying intensities yielded a better understanding of dynamics of rainfall storage on woody vegetation. Branches of eight species
generally retained more water at higher rainfall intensities than at lower intensities, but incremental storage gains decreased as rainfall intensity increased. Leaf area was the best predictor of storage, especially for broadleaved species. Stored water ranged from 0.05
to 1.1 mm effective depth on leaves, depending on species and rainfall intensity. Storage was generally about 0.2 mm greater at rainfall intensity 420 mm h1 than at 20 mm h1. Needle-leaved branches generally retained more water per leaf area than did branches
from broadleaved species, but branches that stored most at lower rainfall intensities tended to accumulate less additional storage at
higher intensities. A simple nonlinear model was capable of predicting both magnitude (good model performance) and temporal
scale (fair model performance) of storage responses to varying rainfall intensities. We hypothesize a conceptual mechanical model
of canopy storage during rainfall that includes the concepts of static and dynamic storage to account for intensity-driven changes in
storage. Scaling up observations to the canopy scale using LAI resulted in an estimate of canopy storage that generally agrees with
estimates by traditional methods.
2005 Elsevier Ltd. All rights reserved.
Keywords: Canopy storage; Rainfall simulation; Canopy interception; Throughfall; Rainfall intensity; Forest canopy
1. Introduction
Detention of water on vegetation is the basic process
controlling interactions of precipitation with plant canopies. Water temporarily stored on canopy surfaces is
readily evaporated and is therefore an important component of the hydrological cycle in most regions. In
addition, temporary storage of rainfall incident on canopies smoothes rainfall intensities and reduces maxi-
*
Corresponding author. Tel./fax: +1 225 578 4169.
E-mail address: [email protected] (R.F. Keim).
0309-1708/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.advwatres.2005.07.017
mum rainfall rates in throughfall compared to rainfall
[65,34].
Canopy storage is difficult to measure in the field.
Several researchers [52,12,5,6] have estimated canopy
storage of entire stands by measuring their attenuation
of electromagnetic waves, Hancock and Crowther [22]
calculated storage on standing trees by measuring
deflections of branches under load from stored water,
while Lundberg [44] and Storck et al. [61] measured canopy storage of snow by weighing entire trees.
All of these methods are expensive and impractical for
many applications, so researchers and practitioners most
often estimate canopy storage indirectly. The running
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
difference between rainfall and the sum of throughfall
and evaporation has sometimes served as an estimate
of storage during storms [47,69], but the most common
approach remains estimating canopy storage ‘‘capacity’’
from storm-total precipitation by the method of Leyton
et al. [40]. This method results in a simple estimate of
water available for evaporation by disregarding details
of the rainfall-storage–drip process.
Despite the general success of the canopy storage
capacity concept as the basis of estimating evaporation
from intercepted water, results of direct [12,35] and indirect [34] field investigations have confirmed assumptions
of the Rutter et al. [58] model and laboratory findings by
Aston [1] that storage increases with rainfall intensity.
Because these storage changes are the mechanism of
intensity smoothing, a better understanding is required
for a process-based model of intensity smoothing by
canopies. The scarcity of data means models have so
far been limited to black-box approaches at the canopy
scale [34]. Ultimately, a physically-based understanding
of canopy storage responses to rainfall would allow
mechanistic modeling of throughfall at fine temporal
and spatial scales. Whelan and Anderson [68] and Davie
and Durocher [16,17] have demonstrated the potential
of such models for predicting evaporation, but did not
consider intensity effects.
An alternative to measuring or estimating storage in
large, complex forest canopies in varying meteorological
conditions is to measure storage on smaller units of canopy under simulated rainfall in a controlled environment. Most previous experiments of this type have
generally been designed to estimate parameters required
by the widely-used canopy interception evaporation
model of Rutter et al. [58]: either canopy storage [28,
31,15,62], or storage-dependent drainage rates [23,55,
56]. There has been only limited work with simulated
rainfall to quantify canopy storage and drainage in
response to rainfall of varying characteristics, most
notably Aston [1], who measured storage and drip on
tree seedlings of varying leaf area under several rainfall
intensities, and Calder et al. [13], who measured storage
in tree branches under simulated rainfall of varying drop
sizes and intensities. However, both of those experiments applied simulated rainfall of temporally constant
intensity, and there have been no published experimental observations of storage responses to temporally varying rainfall intensities.
The objective of this research is to quantify storage of
precipitation on vegetation in response to varying rainfall intensities. First, we will describe the storage responses to varying rainfall intensities of branches from
several species of trees and shrubs of varying morphology. Second, we will generalize and quantify canopy
storage responses to varying rainfall intensities and the
resulting smoothing effect on throughfall intensities by
modeling vegetation as a reservoir for precipitation.
975
Finally, we will estimate canopy-scale storage during
rainfall rates typical of field conditions by scaling up
branch-scale observations.
2. Methods
2.1. Vegetation samples
We collected vegetation samples up to 2 m length by
cutting branches from eight different species of shrubs
and trees common to the Pacific Northwest region of
the USA. We selected five broadleaved species [Acer
macrophyllum (bigleaf maple), Alnus rubra (red alder),
Rubus spectabilis (salmonberry), Rubus parviflorus
(thimbleberry), and Acer circinatum (vine maple)] and
three needle-leaved species [Pseudotsuga menziesii ssp.
menziesii (Douglas-fir), Tsuga heterophylla (western
hemlock), and Thuja plicata (western redcedar)] (nomenclature follows [57]). A. macrophyllum, A. rubra, P. menziesii, T. heterophylla, and T. plicata are overstory trees,
and R. spectabilis, R. parviflorus, and A. circinatum are
woody shrubs. Each sample underwent rainfall simulation less than six hours after collection, and we sealed
the cut end of each sample in paraffin to prevent wilting
during this time. We collected samples and conducted
tests on broadleaved species during August and September of 2002, and needle-leaved species from January to
June of 2003.
Subsequent measurements of the samples resulted in
three ways to characterize each sample: biomass of sample, biomass of leaves/branchlets, and (one-sided, projected) leaf area (LA). After rainfall simulation, we
disassembled each sample into leaves and stems (for
broadleaved species) or branchlets and stems (for needle-leaved species), arranged the resulting parts without
overlap in a frame, and took vertical-view digital photographs of the frame and vegetation for measurement of
surface area. Finally, we obtained the dry biomass of
leaves/branchlets and stems by weighing samples oven
dried at 65 C for at least 24 h.
Analysis of the digital photographs of disassembled
samples entailed using VegMeasurement software
(D. Johnson, Oregon State University Department of
Rangeland Resources) to classify each pixel of images
as either leaf or not leaf, using a qualitatively calibrated
filter to discriminate greenness or brightness, and calculate LA as the proportion of the image classified as
leaves multiplied by the total area of the image. We used
the definition of LA as one-sided, projected leaf area to
approximate leaf surface area exposed to rainfall. This
approximation may have been more accurate for the
broadleaved species than it was for needle-leaved species, which are more round in cross section. Therefore,
we also estimated LA for needle-leaved samples using
published allometric relationships between total surface
976
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
area of the needle-leaved species and projected LA [2,3]
and biomass [54,32], but these methods resulted in
widely varying estimates of total surface area. In addition, surfaces of broad leaves are not flat. For example,
hairs on leaf surfaces are also likely to affect storage, but
are not easily measured. Therefore, we adopted projected LA as measured in the photographs as the
standard for all species. Analysis of stem surface area
was the same as analysis of leaf area, but applying a
correction to projected area to estimate total surface
area assuming cylindrical cross section: surface area =
p Æ (projected area).
2.2. Rainfall simulation and data collection
We constructed a rainfall simulator in the laboratory
that consisted of individually controllable nozzles
employable singly or in groups to simulate four intensi-
Manual Switch (x 5)
110 V AC
Solenoid Valve /
Spray Nozzle
(x 5)
Electronic Balance
Manifold
CPU
City Water
Valve
Aircraft Cable
Pressure Regulator
Vegetation Sample
Fig. 1. Diagram of the experimental apparatus for simulating rainfall
and measuring water stored on sample branches.
ties of rainfall (Fig. 1). The spray rig consisted of a manifold under constant pressure feeding five separate spray
nozzles. Each spray nozzle was mounted on the end of a
valve stub pipe from the manifold, so it could be activated independently of the other nozzles. Each valve
was controlled by an electrical solenoid (similar to the
design of Miller [48]) wired to a wall switch at the rainfall simulator control panel. This design allowed changing of intensity in about 2 s. Water supply and pressure
was by filtered municipal water supply in combination
with a pressure regulator to maintain 1.4 bar (20 psi)
at the nozzles. Pressure gauges between each solenoid
valve and spray nozzle allowed monitoring of pressure
at each nozzle independently, so that head losses in
the system did not affect calibration. The spray rig was
4.9 m above the ground (4 m above the vegetation
sample).
Four different combinations of one or more nozzles allowed simulation of four rainfall intensities: 20, 60, 250
and 420 mm h1 (Table 1). These intensities simulate
moderate to very heavy rain, so that evaporation from
samples was small compared to rainfall and drip rates.
The mean drop sizes produced by each nozzle or combination of nozzles (according to specifications published
by Spraying Systems Co., Wheaton, Illinois, USA)
approximately matched those of natural rainfall (Table
1), although the range of drop sizes was likely less in
spray than in natural rainfall. Calibration was by test
simulations into conterminous 0.25-m2 pans placed in a
grid on the floor beneath the spray. The coefficient of variation of water accumulated in these pans was 0.2–0.4
(depending on intensity) in the middle 2.25 m2 of the sample area where the vegetation samples were placed.
We suspended vegetation specimens on a 5-mm cable
attached to an electronic balance (Mettler-Toledo
SR32001) mounted on the ceiling above the spray rig
(Fig. 1). A computer controlled the balance and recorded the raw weight of the sample at 10 Hz during
spraying. Adjustable cross-member cables on the suspension cable allowed each branch to be suspended at
the same orientation as it was growing in the field. This
position was maintained throughout the test, so that
mechanical responses of the branches to simulated rainfall did not exactly duplicate the natural situation.
Table 1
Rainfall intensities and mean drop diameters produced by the rainfall simulator
Mean rainfall
intensity (mm h1)
Nozzlesa
Mean drop
size (mm)
Expected mean drop
size range (mm)b
20
60
250
420
1 · G3.5 (1.60 mm)
2 · G6.5 (2.39 mm)
G3.5 + 2 · G6.5 + G25 (9.55 mm)
G3.5 + 2 · G6.5 + 2 · G25
1.0
1.3
2.5
2.8
1.1–2.2
1.5–2.6
2.0–3.4
2.3–3.7
a
b
Model numbers for nozzles produced by Spraying Systems, Inc. (Wheaton, Illinois, USA).
Range of values reported by Laws and Parsons [37], Best [4], and Mason and Andrews [45].
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
Each test began with simulated rainfall at the lowest
intensity (20 mm h1), and continued until weight of the
sample reached a quasi-steady value (Ssteady). We then
increased the rainfall intensity to 60 mm h1, waited
for the weight to stabilize, and repeated the sequence
for intensities of 250 and 420 mm h1. When the weight
stabilized at the 420 mm h1 intensity, we turned off the
simulator and continued weighing the sample until drip
rates dropped substantially (3–5 min) and we judged
evaporative losses may have been a large enough proportion of the weight loss to confound estimation of
drip rates.
There were two corrections to the raw weight data required for measurement of water stored during simulated rain. First, we subtracted from the raw data the
time-varying weight of water detained on the suspension
cable, obtained by running tests with no vegetation sample. Second, we subtracted the apparent weight from the
force of water drops impinging on the specimen and suspension cable. Mass accumulating on branches requires
that force be exerted on the branch, caused by momentum of the falling drops being transferred to the branch
[21]. This force varied by rainfall intensity and by vegetation specimen because the orientation of vegetative
surfaces to rainfall and flexibility of the vegetation varied. It was therefore necessary to estimate the force of
raindrops on every sample for each intensity. We accomplished this by analyzing the response of the apparent
sample weight to 10-s interruptions in spray while the
sample was at steady-state storage. For each sample
and intensity, we estimated the force as the instantaneous apparent additional weight caused by the onset
of simulated rainfall at the end of the 10-s interruption,
averaged over four repetitions. The force of raindrops
accounted for less than 5% of the total apparent weight
of all samples, but water stored on the suspension cable
accounted for up to 17% of the apparent weight for samples that were very light.
2.3. Analysis of rainfall-storage–drainage relationships
Drainage (D; [dimensions L3T1]) from a linear reservoir is linearly proportional to storage (S [L3]):
DðtÞ ¼ kSðtÞ;
ð1Þ
where k [T1] is a constant of proportionality. In canopy
storage, the analysis is customarily one-dimensional, so
the dimensions of D become [LT1], and the dimensions
of S become [L].
The equation of mass conservation for water in the
canopy, including rainfall (R; [LT1]), storage, and drip,
is
RðtÞ DðtÞ ¼
dS
.
dt
ð2Þ
977
In Eq. (2), D subsumes stemflow and evaporation; in our
experiments stemflow could not occur and we ignored
evaporation as low compared to R and D. Substituting
Eq. (1) into Eq. (2) gives the governing equation for
mass conservation in a canopy where drip is proportional to storage:
RðtÞ kSðtÞ ¼
dS
.
dt
ð3Þ
Solving Eq. (3) with constant R and initial condition
S(0) = 0 gives
SðtÞ ¼
R
ð1 ekt Þ.
k
ð4Þ
The limit of Eq. (4) as t ! 1 gives
S1 ¼
R
;
k
ð5Þ
which is the storage when steady inflow and outflow approach equilibrium and storage approaches steady state
(S1) [14]. Eq. (4) describes canopy storage approaching
S1 asymptotically from a dry canopy at the onset of
rainfall. This relationship has a long history of application in canopy interception research [39,46], but its
implication for canopy storage during rainfall of varying
intensities has not been fully explored.
Linear reservoirs have the property that storage is linearly related to (steady) rainfall intensities by Eq. (5).
However, the only published data on intensity-dependent storage suggests that incremental gain in steadystate storage decrease with rainfall intensity [1]. Also,
Calder [8] found linear drainage (Eq. (3)) did not fit
experimentally determined recession rates of storage
(R = 0 and S(0) > 0), and most other research in canopy
interception has also assumed or experimentally found
some form of nonlinearity in the storage–drainage relationship (see review by Keim and Skaugset [34]). Therefore, we next consider the possibility that a nonlinear
reservoir is a more appropriate model for canopy
storage.
A simple nonlinear reservoir is
DðtÞ ¼ bSðtÞn ;
ð6Þ
where n is a dimensionless constant of nonlinearity and
b is analogous to k but with dimensions [L1nT1].
Although Eq. (6) is not commonly applied in canopy
storage modeling, Domingo et al. [18] found it described
the observed storage–drainage relationship better than
the more commonly used drip equation originally assumed by Rutter et al. [58]. Substituting Eq. (6) into
Eq. (2) gives the governing equation for mass conservation in a canopy where drip is related to storage by Eq.
(6):
n
RðtÞ bSðtÞ ¼
dS
.
dt
ð7Þ
978
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
Analytical solutions to Eq. (7) are only tractable for
some values of n [60]. However, it is possible to estimate
k and n from data for some special cases. One such case
is during steady-state rainfall at t ! 1, when storage is
also steady (S = S1), dS/dt = 0, and Eq. (7) reduces to
S 1 ¼ cRN ;
a
ð8Þ
N
1N
N
T ] [60].
where N = 1/n [-] and c = (1/b) [L
To assess the viability of reservoir models (Eqs. (5)
and (8)) for modeling steady-state storage under varying
rainfall intensities, we approximated S1 with Ssteady for
each of the four intensities in the simulated rainfall tests,
and then fitted parameters for the linear and nonlinear
models. We calculated parameter k of the linear model
(Eq. (5)) from Ssteady and R of each intensity of all tests.
We fitted parameters of the nonlinear model (Eq. (8)) by
numerical searches for values of c and N that optimized
fit of Eq. (8) to observed Ssteady for each test sequence of
R. The parameter searches maximized the Nash–Sutcliffe efficiency, E, as the objective function [51].
We assessed the ability of the models to predict storage responses to varying rainfall intensity by their ability
to predict the time to steady-state storage, Ssteady, after
the onset of a new intensity. It was difficult to precisely
estimate this time from data, however, because of noise
in the measurements of sample weight and because samples sometimes slowly increased weight during ‘‘steadystate’’ storage (Section 3). To avoid this problem, we
measured response of each sample as the time to reach
storage equal to 0.8 Æ Ssteady after onset of increased
intensity (T80). The variance of fractional response
times is less than the variance of maximum response
times when initial rates of change of responses are faster
than rates of change near the maximum response, so
fractional response variables are more precise estimates
of response times [29].
3. Results
3.1. Storage responses to varying intensities
Leaf area (LA) was the characteristic of samples most
related to steady-state water storage, especially for
broadleaved species (Fig. 2a). Needle-leaved species,
especially T. heterophylla, stored more water per LA
than did broadleaved species at all intensities. Conversely, needle-leaved species stored less water per biomass than did broadleaved species, but biomass of
samples was an unreliable predictor of storage across
species (Fig. 2b). Species with thin leaves or thin stems
(R. spectabilis, R. parviflorus, and A. circinatum) stored
more water per biomass than species with thicker leaves
or larger stems (e.g., needle-leaved species). Biomass of
leaves was of intermediate value as a predictor of storage (Fig. 2c). The highest specific storage in relation to
b
c
Fig. 2. Storage of water on branches under simulated rainfall,
normalized to either (a) leaf area; (b) leaf biomass; or (c) total biomass
of samples. Values are equilibrium storage averaged across all tests for
each species (5 6 n 6 10).
leaf biomass was on the thin, rough leaves of Rubus
spp., especially at the highest simulated rainfall intensities. The effectiveness of LA in normalizing storage
among species also extended to effectiveness in normalizing among samples of the same species (not shown).
Therefore, we conducted further analyses on the basis
of storage (g) per LA (cm2). Conveniently, 10 Æ storage/LA = average depth of water on leaf surfaces
(mm), and storage (mm) Æ leaf area index (LAI; dimensionless) = canopy storage per LAI (mm).
Steady-state storage generally increased with increased simulated rainfall intensity (Table 2, Figs. 2
and 3a). Only three of 57 samples (A. rubra 5, A. rubra
8, and R. parviflorus 7) stored less water at rainfall intensity 60 mm h1 than at 20 mm h1, and one sample
(T. heterophylla 6) stored less water at 250 mm h1 than
at 60 mm h1. All 57 samples stored more water at
420 mm h1 than at 250 mm h1 (Table 2).
The typical response of most species and samples to
the onset of simulated rainfall at 20 mm h1 was for
water stored on branches to come to equilibrium within
5–10 min, with equilibrium storage reached much more
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
979
Table 2
Sample branches subjected to simulated rainfall
Sample
Position in canopya
Branch
Acer macrophyllum (bigleaf maple)
1
L
2
L
3
L
4
L
5
L
6
L
7
L
Leaf area (m2)
Dry weight (g)
Leaves only
Steady-state storage (g) at intensity (mm h1)
20
60
250
420
49.2
86.7
38.3
82.0
66.8
99.1
62.7
18.9
41.6
20.3
27.9
29.2
42.7
25.2
0.56
0.63
0.51
0.89
0.63
0.83
0.79
67
141
106
183
134
165
74
95
157
125
191
136
191
106
139
181
160
220
175
243
184
201
206
170
237
188
279
225
67.5
41.3
39.2
59.9
28.2
91.2
37.5
123.2
80.6
60.3
15.1
20.5
16.2
17.6
11.8
42.0
11.5
60.7
36.7
25.3
0.31
0.25
0.30
0.34
0.22
0.56
0.23
0.70
0.60
0.29
54
34
54
94
59
81
40
184
80
54
68
37
56
99
57
99
47
182
88
66
103
67
79
110
69
160
62
199
118
109
121
76
94
123
75
190
81
207
135
122
Rubus spectabilis (salmonberry)
1
S
2
S
3
S
4
S
5
S
6
S
7
S
8
S
9
S
10
S
28.2
14.6
25.1
14.9
31.3
32.5
57.2
27.6
22.9
11.4
12.7
6.5
10.1
6.4
13.9
11.6
22.6
11.3
8.7
3.4
0.54
0.25
0.42
0.25
0.44
0.47
0.67
0.21
0.15
0.17
50
29
85
30
95
81
162
26
25
24
66
45
109
37
116
94
173
33
35
32
122
58
144
72
169
131
214
74
62
46
148
69
155
86
179
150
231
75
69
52
Rubus parviflorus (thimbleberry)
1
S
2
S
3
S
4
S
5
S
6
S
7
S
28.2
25.1
17.5
14.2
38.4
24.9
24.5
13.8
11.8
7.7
4.7
16.8
12.0
9.4
0.40
0.38
0.26
0.19
0.35
0.32
0.30
46
30
30
32
24
19
35
59
39
34
39
36
24
34
120
88
58
63
71
57
57
153
120
81
77
114
85
76
Acer circinatum (vine maple)
1
S
2
S
3
S
4
S
5
S
6
S
62.5
32.7
32.5
76.2
82.4
61.2
14.7
9.1
8.1
16.3
25.6
15.1
0.63
0.37
0.40
0.63
0.96
0.57
118
80
71
102
145
123
150
81
82
123
185
133
183
93
110
165
237
158
193
99
119
190
270
176
Pseudotsuga menziesii ssp. menziesii (Douglas-fir)
1
L
134.8
2
L
260.0
3
U
208.5
4
L
132.7
5
U
128.1
6
L
109.7
88.8
193.9
118.2
110.6
83.4
87.8
1.11
1.33
0.44
0.94
0.44
0.83
253
360
161
186
132
157
273
415
181
203
145
209
295
525
223
261
172
241
323
570
248
284
190
259
Tsuga heterophylla (western hemlock)
1
L
2
U
3
L
4
L
120.8
37.2
52.8
14.1
0.30
0.16
0.48
0.13
165
71
223
68
217
87
263
87
289
109
314
104
(continued
323
119
323
116
on next page)
Alnus rubra (red alder)
1
L
2
M
3
L
4
L
5
L
6
L
7
L
8
M
9
U
10
M
141.6
54.9
108.7
32.3
980
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
Table 2 (continued)
Sample
Position in canopya
Branch
5
6
L
L
Thuja plicata (western redcedar)
1
L
2
L
3
M
4
M
5
L
a
Leaf area (m2)
Dry weight (g)
Leaves only
Steady-state storage (g) at intensity (mm h1)
20
60
250
420
85.8
54.3
66.6
40.2
0.29
0.19
114
92
134
100
156
99
170
105
196.3
146.1
108.2
48.6
217.3
127.1
113.4
83.6
33.5
127.1
0.75
0.64
0.23
0.23
0.64
191
109
76
59
182
243
153
90
74
232
323
200
116
97
282
354
223
127
104
305
L = lower, M = middle, U = upper canopy; S = shrub layer.
a
b
samples accumulated storage monotonically until reaching equilibrium, but eight of the 57 samples reached an
early peak in storage followed by lower equilibrium storage (Fig. 3b). The minimum storage reached during successive 10-s interruptions in spraying increased or
remained constant with increasing rainfall intensity for
all but 16 of the 57 samples (Fig. 3c). The final common
atypical behavior was slow accumulation of storage
(Fig. 3c); variance of response times is discussed in Section 3.2.
3.2. Modeling storage responses to varying intensities
c
Estimates of the linear reservoir parameter k from (5)
varied by species and simulated rainfall intensity
(Fig. 4). Differences by species are the result of varying
equilibrium storage (Fig. 2). Differences in the estimate
of k by intensity, however, show that the linear model
cannot adequately describe the rainfall-storage–drip
relationship for the sample branches. This lack of fit is
also evident in plots of storage as a function of rainfall
1.6
k
1.2
Fig. 3. Example data from tests of rainfall simulation on branches. In
each case, the black line is the depth-equivalent mass of water stored
on a branch, thin vertical lines indicate step increases in rainfall
intensity (shown in panel c). Dips in stored water are the result of
10-s interruptions in sprinkling. Data are (a) R. spectabilis sample 5;
(b) R. parviflorus sample 5; (c) and T. heterophylla sample 6.
0.8
0.4
quickly after the onset of successively higher intensities
(Fig. 3a). However, there were several atypical but repeated behaviors worthy of note. Nineteen of the 57
samples continued slow increases in storage throughout
at least one intensity (Fig. 3b). For later analyses based
on steady-state storage for samples that showed this
behavior, we used the maximum storage near the end
of the test. Starting from an initially dry condition, most
0.0
0
100
200
Rainfall (mm h-1)
300
400
Fig. 4. Constant of proportionality, k, of a linear reservoir model
describing water stored on branches, estimated from equilibrium
storage and intensity of simulated rainfall for each of eight species (see
key in Fig. 2). Values are normalized by one-sided sample leaf area and
averaged across all samples for each species.
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
981
0.8
Fractional 10-s Drainage, Model
Storage/Leaf Area (mm)
0.8
0.6
0.4
0.2
0.0
0
100
200
300
400
500
Rainfall (mm h-1)
0.7
0.6
e
nt
ll I
fa
0.4
n
ai
R
0.3
0.2
0.1
0.0
0.0
Fig. 5. Nonlinear models of water stored on branches of several
species at rainfall-intensity-dependent equilibrium, fitted to data
averaged across all tests per species (see key in Fig. 2).
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Fractional 10-s Drainage, Data
Fig. 6. Observed drainage of water stored on branches during 10-s
interruptions of simulated rainfall, expressed as fractions of intensitydependent equilibrium storage, compared to predictions of a nonlinear
rainfall-storage model calibrated to equilibrium storage amounts.
Observed values are averages of 1–4 interruptions per sample, and
averaged across all samples for each species (5 6 n 6 10). The diagonal
line indicates 1:1 agreement between data and models.
intensity (Fig. 2), which must be linear and pass through
the origin for proper fit of a linear model.
The nonlinear model (Eq. (8)) fit observations of
intensity-dependent storage well (Fig. 5). The efficiency
of best-fit models predicting storage of individual samples over varying simulated rainfall intensity ranged
from E = 0.76 to E > 0.99, with mean E = 0.97. Efficiency of the model in predicting average responses by
species ranged from E = 0.97 to E > 0.99, with mean
E = 0.99.
In addition to efficiently predicting equilibrium storage (the calibration data), the fitted nonlinear models
also predicted responses to varying rainfall intensity surprisingly well. The fitted models predictions of fractional drainage in response to the 10-s interruptions in
spraying matched observed values for some species,
but underestimated fractional drainage from R. spectabilis and R. parviflorus (Fig. 6). The fitted models
1000
1000
a
T80 (sec), Model
ity
ns
0.5
c
100
100
10
10
1
1
0
0
1
0
10
100
1000
1000
0
1
10
100
1000
1
10
100
1000
1000
b
d
100
100
10
10
1
1
0
0
0
1
10
100
1000
0
T80 (sec), Data
Fig. 7. Relationship between observed and modeled response times of water stored on branches to increases in intensity of simulated rainfall,
quantified by the time from onset of rainfall intensity to 0.8 Æ eventual equilibrium storage at the new intensity (T80). The panels are for simulated
rainfall intensity steps from (a) 0 to 20 mm h1, (b) 20 to 60 mm h1, (c) 60 to 250 mm h1, and (d) 250 to 420 mm h1. Each symbol represents the
response time of a single sample (see species key in Fig. 2), and lines indicate 1:1 agreement between data and models.
982
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
predicted the increases of 10-s fractional drainage with
rainfall intensity well, as evidenced by the slope of unity
in a plot of model vs. data (Fig. 6).
The nonlinear model underpredicted the time required for samples to come to equilibrium storage
(T80) for more than 80% of samples, and there was considerably more variance in observations compared to
modeled response times (Fig. 7). Variation in measured
T80 was not clearly related to any variable except rainfall
intensity and species; i.e., sample leaf area and biomass
were not significant predictors of response time. Needle-leaved species were slower than broadleaved species
to accumulate storage at the onset of sprinkling and in
response to intensity increases up to 25 mm h1, but this
relationship was reversed at the highest intensity (Fig. 7;
Table 2). The species that was fastest to respond to all
new intensities was A. rubra. Overall, fast response times
were generally associated with smooth, broad leaves, and
complex vegetative surfaces were slower to respond.
Best-fit parameters of the nonlinear model varied
consistently by species and vegetation morphology,
and best fits were in a distinct region of parameter space
(Fig. 8). Needle-leaved species, especially T. heterophylla, tended to be best described by larger values of c than
were the broadleaved species, which corresponds to generally higher equilibrium storage amounts per leaf area
(Figs. 2 and 5). Broadleaved species characterized by
low c were also characterized by higher N, which indicates more nearly linear increases in storage with intensity (N = 1 is a linear response). Physically, this means
that branches that stored more water at lower rainfall
intensities tended to accumulate proportionally less
water at higher rainfall intensities (on a per-leaf-area
basis); species that stored water more effectively were
less sensitive to variations of intensity.
Least-squares regression between best-fit parameters
of the nonlinear model from all tests yielded c =
1
Needle-leaved:
c = 0.39e -4.4N
R 2 = 0.24
c
0.1
0.01
0.001
0.0
Broadleaved:
c = 0.31e -5.9N
R 2= 0.96
0.2
0.4
0.6
N
Fig. 8. Relationship between the parameters of a nonlinear model
describing storage on branches as a function of simulated rainfall
intensity (see species key in Fig. 2).
0.41e6.5N with R2 = 0.87. However, the regression for
the broadleaved species alone (c = 0.31e5.9N; R2 =
0.96) was much better than for needle-leaved species
alone (c = 0.39e4.4N; R2 = 0.24) (Fig. 8). The consistent
relationship between c and N suggests that dynamic
storage on broadleaved species as a group depends on
fewer variables than on needle-leaved species, perhaps
because of less intraspecific morphological variability.
4. Discussion
Measurements of specific water storage in this study
of 0.10–0.46 mm (up to 0.76 mm for T. heterophylla),
depending on species and rainfall intensity (Table 2,
Fig. 2), are comparable to previous direct measurements. Grah and Wilson [20] measured equilibrium
water storage on Pinus radiata during heavy spray at
0.11 mm and on Baccharis pilularis at 0.25 mm. Aston
[1] measured storage after 2 min of drainage from seedlings of two eucalypts, Acacia longifolia, and P. radiata,
at 0.03–0.18 mm. We estimate these values correspond
to about 0.12–0.36 mm during rainfall, based on drip
rates measured in our study. Herwitz [27] measured
equilibrium water storage on saplings of Australian
rainforest species of 0.23–0.32 mm under simulated
heavy rainfall.
There is a discrepancy in specific storage measured by
dipping vegetation into water compared to sprinkling
experiments, especially when water is blotted from foliage to simulate windy conditions. Storage measured this
way is nearly an order of magnitude lower than under
rainfall simulation. Using this method, Crockford and
Richardson [15] measured storage at 0.017–0.028 mm
for three eucalypts, Liu [42] measured storage at
0.036–0.041 mm for needle-leaved and broadleaved species, and Llorens and Gallart [43] measured storage at
0.104 mm (0.043 mm when blotted) for Pinus sylvestris
needles.
Applying these estimates to canopy scale is possible
by multiplying leaf storage (mm) by leaf area index
(LAI; dimensionless). For example, results of our experiment suggest branch storage in Douglas-fir at constant
rainfall intensity of 20 mm h1 would be S = cRN =
0.23 Æ 200.13 = 0.34 mm. In a pure Douglas-fir forest
with LAI of 7, canopy storage would be 0.34 Æ 7 =
2.4 mm, ignoring storage on woody biomass. For
comparison, Link et al. [38] estimated saturation storage
at 3.0–4.1 mm for the Wind River Canopy Crane forest,
which has LAI 6.9 in mixed needle-leaved species,
including Douglas-fir [63], and a major component of
woody biomass in the canopy [53]. Because the sprinkling method of branch-scale storage appears to match
field estimates, it appears that the consistently lower estimates of storage obtained by the dipping method are
inaccurate.
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
The poorly defined nature of canopy storage terms is
a source of confusion that leads to conceptual and
methodological difficulties in canopy interception research. For example, the concept of canopy storage
‘‘capacity’’ is common in the field, but definitions of this
state vary tremendously. For example, Rutter et al. [58]
and earlier workers (e.g., Leyton et al. [40]) defined storage capacity as ‘‘saturation’’ in terms of maximum availability for evaporation, recognizing that actual retention
may exceed this value. Later modelers simplified the
concept to a bucket model, defining ‘‘capacity’’ as a single value above which no more precipitation is retained
[50]. The rationale for this simplification generally has
been that water in excess of canopy saturation is not
available for evaporation and need not be considered
[41]. Another common definition of canopy storage
capacity includes the storage amount when drip stops
[26]. Confounding all these definitions are considerations of rainfall intensity and wind, which have both
generally been assumed to reduce storage ‘‘capacity’’
[25,9,19].
Resolution of this confusion depends on adoption of
a physically-based conceptual model of canopy storage.
We suggest here a conceptual mechanical model of rainfall passage into, on, and through canopies. The goal is
to develop a consistent nomenclature and define the processes in physical terms.
The kinetics of water detained on branches during
rainfall entails static interfacial forces at the contact between water and vegetation and the atmosphere, as well
as dynamic forces arising from motion of water in and
impinging upon the canopy. The traditional notion of
canopy storage capacity approximately coincides with
the conceptual condition where the amount of water
present on canopy surfaces is such that gravitational
forces are balanced by interfacial and normal forces
resisting movement of water, and all dynamic forces
are zero. Rain falling on a canopy in this state decelerates upon impact, decreasing its momentum. This transfer of momentum has two relevant effects on water
storage within the canopy. First, it adds net downward
dynamic forces that work with gravity to favor drip.
Canopy interception researchers have long recognized
this effect. For example, Calder et al. [13] found that
holding intensity approximately steady while increasing
momentum of rainfall by increasing drop size results in
lower storage. The second effect of rainfall transferring
momentum to the canopy is to add mass of water [21],
at least temporarily.
While evaporation loss from canopies depends on the
reservoir of stored water in the canopy, probably largely
retained by static forces (‘‘static canopy storage’’), the
intensity smoothing effect of canopies depends more
upon dynamic forces (‘‘dynamic canopy storage’’). Specifically, we hypothesize dynamic canopy storage (and
hence intensity smoothing) to be mainly the result of
983
the balance between the two main effects of rain falling
on wet canopies: addition of water and dislodging of
existing storage. Smoothing is probably enhanced by
refilling of static storage dislodged by dynamic forces
external to the rainfall (i.e., wind; [30,38]), as well as
by momentum transferred out of canopy water by viscous flow along vegetative surfaces [28].
An analogy for illustrating the concepts of static and
dynamic storage is a porous medium. We envision ‘‘static storage capacity’’ as analogous to field capacity of a
soil. This wetness is reached relatively quickly after rainfall ceases, and is an important characteristic of the
medium. Water in excess of static storage capacity (field
capacity) drains quickly, but the drainage is not instantaneous. The difference between the porous media that
are soil and canopies is that water movement in soil
can usually be approximated by potential flow and
momentum neglected at all time scales of measurement,
but momentum of water in canopies appears to be an
important control on throughfall generation at short
timescales.
The balance of forces and supply of mass that affects
dynamic storage can conceptually result in net drainage
of stored water, as assumed by many previous authors
[10,11], but our experiments and the experiments of
Aston [1] suggest that the more common result is net
increased storage. The variability of behavior among
species and within species in this study makes it difficult
to infer what vegetation characteristics control the
balance of forces. However, roughness of leaves seems
less important than roughness of the branch as a whole.
For example, species such as R. parviflorus with rough,
broad leaves were much less effective at storing
water than species with smooth leaves but complexity
at the branch scale in the form of small needles as in
T. heterophylla (Fig. 5). Other potentially important factors are leaf shape, epiphytes, and surface paniculate
matter.
It is unclear why slow accumulation of storage occurred in some but not all samples in this experiment;
we were unable to determine a relationship between
this phenomenon and any sample characteristics except
that none of the samples were of needle-leaved species.
We hypothesize that slow addition of storage may have
been the result of slow accumulation of water on parts
of the samples not directly exposed to the simulated
rainfall, either by drop splash or flow along surfaces
[28].
We were also unable to determine a relationship between sample characteristics and the phenomenon of
nonmonotonic accumulation of storage (Fig. 3b) except
that none of the samples were of needle-leaved species.
We hypothesize that the decrease from the initial maximum storage was the result of storage exceeding some
threshold mass where gravity overcame surface tension
in much the same way that momentum of drops moving
984
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
along surfaces can effect drainage rates that exceed rates
expected by static forces alone in soils [64].
The minimum storage reached during successive 10-s
interruptions in spraying increased or remained constant
with increasing rainfall intensity for all but 16 of the 57
samples (e.g., Fig. 3). We infer that 10 s was too short
for full drainage of dynamic canopy storage in cases
where minimums increased with storage. For the samples where these minimum storage amounts were lower
at higher rainfall intensity (e.g., Fig. 3c), we hypothesize
that the greater momentum imparted by rainfall effected
drainage below levels expected by static forces alone, as
in the case of nonmonotonic approaches to equilibrium
storage.
Because biomass of branches did not correlate well to
rainfall storage (even within the same species), measurements of canopy biomass, while easier to obtain than
surface area, are insufficient for predicting canopy-scale
storage. Unfortunately, this makes it difficult to compare our data to measurements of whole-canopy storage
by Hancock and Crowther [22] and Calder and Wright
[12], who reported only canopy biomass. Water storage
on woody components of canopies can also vary substantially by species and morphology [15], which complicates the relationship between canopy biomass and
storage.
The results from this experiment agree with data presented by Aston [1], who found canopy storage filled faster at higher rainfall intensities than at lower intensities.
This effectively refutes Calders [9] hypothesis that canopy storage increases more quickly at lower intensities.
Although the rainfall simulation work by Calder et al.
[13] appears to support Calders [9] hypothesis, the rainfall simulator used by Calder et al. [13] varied only drop
size substantially, while rainfall intensity remained relatively constant (drop diameter 0.5 and 1.2 mm at R =
36 mm h1 and drop diameter 5.2 mm at R =
45 mm h1). That experiment, therefore, seems to have
been most fitted to determining the simple effect of drop
kinetic energy on storage.
The measurements of storage responses to rainfall
intensity variations in this study are strictly only applicable to branches receiving rainfall instead of drip from
branches above, because the drop size distribution of
drip does not match that of rainfall [67,7,24]. Canopy
drip tends to be composed of larger drops than rain,
so that kinetic energy of throughfall is greater than rainfall if drops achieve sufficient velocity [49,67]. Therefore,
understory branches may show less additional storage
with increased intensity, and canopy-scale storage responses to varying rainfall intensity may therefore be
muted compared to branch-scale responses. However,
dense canopies, where drip does not fall far enough to
reach terminal velocity (about 4.5 m for 1-mm drops
and 9 m for 5-mm drops [36]), may act more like large
branches. Also, we speculate that intense rainfall may
be more likely to initiate stemflow or cause accumulation on normally dry part of the canopy by drop splash
[28]. These processes would tend to increase storage by
partitioning water into slower pathways of drainage.
The nonlinear model to describe storage as a function
of rainfall intensity produced generally good fits to
data, and values of the parameters were interpretable
and at least moderately repeated among samples of the
same species and species groups. It would be possible
to derive more sophisticated models, either from first
principles or simply statistically to produce better fits
to data where the simple model we used here was less
than satisfactory. Because storage behavior was fairly
predictable, gains in modeling processes at the branch
scale seem likely to yield insights into storage processes
at the canopy scale.
Davie and Durocher [16,17] presented a model of forest canopies that explicitly represents inflow, storage,
and drainage for each cell of a pixilated canopy.
Although their goals were to reproduce spatial patterns
of throughfall and they represented water movement
through each pixel only stochastically, it would theoretically be possible to use their framework with the results
of our experiments and three-dimensional data of leaf
area and species to parameterize a discrete-element
model of water transfer through canopies. Difficulties
may arise from processes that do not scale well between
the branch scale and canopy scale. Examples include
wind, stemflow, and storage in woody biomass.
5. Conclusions
The results of these rainfall simulation experiments
give quantitative estimates of how storage varies by
rainfall intensity, and suggests that morphological characteristics of vegetation may play a role in this process.
Branches of all eight tested species generally retained
more water at higher rainfall intensities than at lower
intensities, but needle-leaved branches generally retained
more water per leaf area than did branches from broadleaved species. Incremental storage gains decreased as
rainfall intensity increased. Leaf area was the best predictor of storage, but this relationship was stronger for
broadleaved species than for needle-leaved species.
Branches that stored most water at lower rainfall intensities tended to accumulate less additional storage at
higher rainfall intensities. A simple nonlinear model,
S = cRN, was capable of predicting both magnitude
(good model performance) and temporal scale (fair
model performance) of storage responses to varying
rainfall intensities. We hypothesize these processes are
controlled by a conceptual mechanical model of canopy
storage during rainfall that includes the concepts of static storage and dynamic storage to account for intensitydriven changes in storage. Scaling up observations to the
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
canopy scale, using published estimates of LA1 of a
well-studied canopy, resulted in estimates of canopy
storage that generally agreed with estimates obtained
by the traditional Leyton method. The simulated rainfall intensities in this research were high; more work remains to quantify intensity-dependant storage under
lower intensities.
Acknowledgements
This research was supported by grant 00-34158-8978
from the US Department of Agriculture, Cooperative
State Research Education and Extension Service, Centers for Wood Utilization Research. We thank Nalini
Nadkarni and Bob Van Pelt for sharing their Cedar
Flats research plot, and four anonymous reviewers for
helpful comments.
References
[1] Aston AR. Rainfall interception by eight small trees. J Hydrol
1979;42:383–96.
[2] Barclay HJ. Conversion of total leaf area to projected leaf area in
lodgepole pine and Douglas-fir. Tree Physiol 1998;18:185–93.
[3] Barclay HJ, Goodman D. Conversion of total to projected leaf
area index in conifers. Can J Botany 2000;78:447–54.
[4] Best AC. The size distribution of raindrops. Quart J Roy Meteor
Soc 1950;16:16–36.
[5] Bouten W, Swart PJF, de Water E. Microwave transmission, a
new tool in forest hydrological research. J Hydrol 1991;124:
199–230.
[6] Bouten W, Schaap MG, Aerts J, Vermetten AWM. Monitoring
and modelling canopy water storage amounts in support of
atmospheric deposition studies. J Hydrol 1996;81:305–21.
[7] Brandt CJ. The size distribution of throughfall drops under
vegetation canopies. Catena 1989;16:507–24.
[8] Calder IR. A model of transpiration and interception loss from a
spruce forest in Plynlimon, central Wales. J Hydrol 1977;33:
247–65.
[9] Calder IR. A stochastic model of rainfall interception. J Hydrol
1986;89:65–71.
[10] Calder IR. Dependence of rainfall interception of drop size: 1.
Development of the two-layer stochastic model. J Hydrol 1996;
185:363–78.
[11] Calder IR. Rainfall interception and drop size—development and
calibration of the two-layer stochastic interception model. Tree
Physiol 1996;16:727–32.
[12] Calder IR, Wright IR. Gamma ray attenuation studies of
interception from Sitka spruce: some evidence for an additional
transport mechanism. Water Resour Res 1986;22:409–17.
[13] Calder IR, Hall RL, Rosier PTW, Bastable HG, Prasanna K.
Dependence of rainfall interception of drop size: 2. Experimental
determination of the wetting functions and two-layer stochastic
model parameter for five tropical tree species. J Hydrol 1996;185:
379–88.
[14] Chow VT. Handbook of applied hydrology. New York: McGraw-Hill; 1998.
[15] Crockford RH, Richardson DP. Partitioning of rainfall in a
eucalypt forest and pine plantation in southeastern Australia III:
determination of the canopy storage capacity of a dry sclerophyll
eucalypt forest. Hydrol Process 1990;4:157–67.
985
[16] Davie TJA, Durocher MG. A model to consider the spatial
variability of rainfall partitioning within deciduous canopy. I.
Model description. Hydrol Process 1997;11:1509–23.
[17] Davie TJA, Durocher MG. A model to consider the spatial
variability of rainfall partitioning within deciduous canopy. II.
Model parameterization and testing. Hydrol Process 1997;11:
1525–40.
[18] Domingo F, Sánchez G, Moro MJ, Brenner AJ, Puigdefábregas J.
Measurement and modelling of rainfall interception by three semiarid canopies. Agric Forest Meteorol 1998;91:275–92.
[19] Dunkerly D. Measuring interception loss and canopy storage in
dryland vegetation: a brief review and evaluation of available
research strategies. Hydrol Process 2000;14:669–78.
[20] Grah RF, Wilson CC. Some components of rainfall interception.
J Forest 1944;42:890–8.
[21] Grubin C. Mechanics of variable mass systems. J Franklin I
1963;276:305–12.
[22] Hancock NH, Crowther JM. A technique for the direct measurement of water storage on a forest canopy. J Hydrol 1979;41:
105–22.
[23] Hall RL. Further interception studies of heather using a wetsurface weighing lysimeter system. J Hydrol 1985;81:193–
210.
[24] Hall RL, Calder IR. Drop size modification by forest canopies:
measurements using a disdrometer. J Geophys Res 1993;98:
18465–70.
[25] Halldin S, Grip H, Perttu K. Model for energy exchange of a pine
forest canopy. In: Halldin S, editor. Comparison of forest water
and energy exchange models. Copenhagen: International Society
for Ecological Modelling; 1979. p. 59–75.
[26] Hashino M, Yao H, Yoshida H. Studies and evaluations on
interception processes during rainfall based on a tank model. J
Hydrol 2002;255:1–11.
[27] Herwitz SR. Interception storage capacities of tropical rainforest
canopy trees. J Hydrol 1986;77:237–52.
[28] Herwitz SR. Raindrop impact and water flow on the vegetative
surfaces of trees and the effects on stemflow and throughfall
generation. Earth Surf Proc Landf 1987;12:425–32.
[29] Hjerdt KN. Deconvoluting the hydrologic response of a small till
catchment: spatial variability of groundwater level and quality in
relation to streamflow. Ph.D. Dissertation, State University of
New York, College of Environmental Science and Forestry,
Syracuse; 2002. 204 p.
[30] Hörmann G, Brandig A, Clemen T, Herbst M, Hinrichs A,
Thamm F. Calculation and simulation of wind controlled canopy
interception of a beech forest in Northern Germany. Agric Forest
Meteorol 1996;79:131–48.
[31] Hutchings NJ, Milne R, Crowther JM. Canopy storage capacity
and its vertical distribution in a Sitka spruce canopy. J Hydrol
1998;104:161–71.
[32] Ishii H, Ford ED, Boscolo ME, Manriquez AC, Wilson ME,
Hinckley TM. Variation in specific needle area of old-growth
Douglas-fir in relation to needle age, within-crown position and
epicormic shoot production. Tree Physiol 2002;22:31–40.
[34] Keim RF, Skaugset AE. A linear system quantification of
dynamic storage of precipitation on forest canopies. Water
Resour Res 2004;40:W05208. doi:10.1029/2003WR002875.
[35] Klaassen W, Bosveld F, de Water E. Water storage and
evaporation as constituents of rainfall interception. J Hydrol
1998;212–213:36–50.
[36] Laws JO. Measurements of the fall velocity of waterdrops and
raindrops. Trans Am Geophys Un 1941;22:709–21.
[37] Laws JO, Parsons DA. The relation of raindrop-size to intensity.
Trans Am Geophys Un 1943;24:452–60.
[38] Link TE, Unsworth M, Marks D. The dynamics of rainfall
interception by a seasonal temperate rainforest. Agric Forest
Meteorol 2004;124:171–91.
986
R.F. Keim et al. / Advances in Water Resources 29 (2006) 974–986
[39] Linsley RK, Kolher MA, Paulhus JL. Applied hydrology. New
York: McGraw-Hill; 1949.
[40] Leyton L, Reynolds ERC, Thompson FB. Rainfall interception in
forest and moorland. In: Sopper WE, Lull HW, editors. Forest
hydrology. Oxford: Pergamon; 1967. p. 163–78.
[41] Liu S. A new model for the prediction of rainfall interception in
forest canopies. Ecol Model 1997;99:151–9.
[42] Liu S. Estimation of rainfall storage capacity in the canopies of
cypress wetlands and slash pine uplands in north-central Florida.
J Hydrol 1998;207:32–41.
[43] Llorens P, Gallart F. A simplified method for forest water storage
capacity measurement. J Hydrol 2000;240:131–44.
[44] Lundberg A. Evaporation of intercepted snow—review of existing
and new measurement methods. J Hydrol 1993;151:267–90.
[45] Mason BJ, Andrews JB. Drop-size distributions from various
types of rain. Quart J Roy Meteor Soc 1960;86:346–53.
[46] Merriam RA. A note on the interception loss equation. J Geophys
Res 1960;65:3850–1.
[47] Massman WJ. The derivation and validation of a new model for
the interception of rainfall by forests. Agric Meteorol 1983;28:
261–86.
[48] Miller WP. A solenoid-operated, variable intensity rainfall simulator. Soil Sci Soc Am J 1987;51:832–4.
[49] Mosley MP. The effect of a New Zealand beech forest canopy on
the kinetic energy of water drops and on surface erosion. Earth
Surf Proc Landf 1982;7:103–7.
[50] Mulder JPM. Simulating interception loss using standard meteorological data. In: Hutchison BA, Hicks BB, editors. The forest–
atmosphere interaction. Dordrecht, The Netherlands: D. Reidel;
1985. p. 177–96.
[51] Nash JE, Sutcliffe JV. River flow forecasting through conceptual models, I, A discussion of principles. J Hydrol 1970;10:
282–90.
[52] Oszyczka B, Crowther JM. The application of gamma-ray
attenuation to the determination of canopy mass and canopy
water storage. J Hydrol 1981;49:355–68.
[53] Parker GG, Harmon ME, Lefsky MA, Chen JQ, Van Pelt R, Weis
SB, et al. Three-dimensional structure of an old-growth
Pseudotsuga–Tsuga canopy and its implications for radiation
balance, microclimate, and gas exchange. Ecosystems 2004;7:
440–53.
[54] Pike LH, Rydell RA, Denison WC. A 400-year-old Douglas fir
tree and its epiphytes: biomass, surface area, and their distributions. Can J Forest Res 1977;7:680–99.
[55] Pitman JI. Rainfall interception by bracken in open habitats—
relations between leaf area, canopy storage and drainage rate. J
Hydrol 1989;105: 317–34.
[56] Pitman JI. Rainfall interception by bracken litter—relationship
between biomass, storage and drainage rate. J Hydrol 1989;111:
281–91.
[57] Pojar J, MacKinnon A. Plants of the Pacific Northwest
Coast. Redmond (WA, USA): Lone Pine; 1984.
[58] Rutter AJ, Kershaw KA, Robins PC, Morton AJ. A predictive
model of rainfall interception in forests, 1. Derivation of the
model from observations in a plantation of Corsican pine. Agric
Meteorol 1971;9:367–84.
[60] Singh VP. Hydrologic systems: vol. 1. Rainfall-runoff modeling. Englewood Cliffs (NJ, USA): Prentice-Hall; 1988.
[61] Storck P, Lettenmaier DP, Bolton SM. Measurement of snow
interception and canopy effects on snow accumulation and melt in
a mountainous maritime climate, Oregon, United States. Water
Resour Res 2002;38:1223. doi:10.1029/2002WR001281.
[62] Teklehaimanot Z, Jarvis PG. Direct measurement of evaporation
of intercepted water from forest canopies. J Appl Ecol 1991;28:
603–18.
[63] Thomas SC, Winner WE. Leaf area index of an old-growth
Douglas-fir forest estimated from direct structural measurements
in the canopy. Can J Forest Res 2000;30:1922–30.
[64] Torres R, Alexander LJ. Intensity-duration effects on drainage:
column experiments at near-zero pressure head. Water Resour
Res 2002;38:1240. doi:10.1029/2001WR001048.
[65] Trimble Jr GR, Weitzman S. Effect of a hardwood canopy on
rainfall intensities. Trans Am Geophys Un 1954;35:226–34.
[67] Vis M. Interception, drop size distributions and rainfall kinetic
energy in four Colombian forest ecosystems. Earth Surf Proc
Landf 1986;11:591–603.
[68] Whelan MJ, Anderson JM. Modelling spatial patterns of
throughfall and interception loss in a Norway spruce (Picea
abies) plantation at the plot scale. J Hydrol 1996;186:335–54.
[69] Xiao Q, McPherson EG, Ustin SL, Grismer ME. A new approach
to modeling tree rainfall interception. J Geophys Res 2000;105:
29173–88.